Soluna Holdings, Inc. (SLNH) Earnings Call Transcript & Summary

Hi. Hello, everyone, and welcome to the water tower fireside chat series. My name is Graham Mattison, one of the publishing analysts here at Water Tower and our sustainability and industrials practice. We're excited to have with us today John Belizaire, Soluna Holdings. I should remention that Soluna's safe harbor statements can be found on their website and in their public filings. As a reminder, this call in all of Water Tower's research is available to all on our website. We will be taking questions from the audience. Those can be submitted via the chat function on the webcast. Hopefully, we'll get to all of them, but our apologies if time does not allow.

So with that, John, great to have you back. And first question is, earlier this month, there was a PR out about the carbon footprint of your data centers. Can you give us a quick summary of what that report said and give us some context as to why it's important and what it means for you?

Hey Graham. Thanks. Happy to do so. And thanks for having me back on these fireside chats, I always enjoy them. Happy New Year, and I hope you had a great holiday. I think it's still safe to be saying those things now. So yes, we did put out an update recently that included a -- I would say, a really important announcement around the carbon footprint of our data centers. It was focused around the Dorothy project. And essentially, what we did is we hired Resurety, they are a software and solutions provider that are dedicated to accelerating the world's transition to zero carbon futures and they've analyzed Soluna's carbon footprint using the metric of something called Local (sic) [ Locational ] Marginal Emissions or LME. And LME is really a measure of the tons of carbon emissions displaced by 1 megawatt of clean energy. So by way of example, if you if you put 1 megawatt on the grid at a specific location, and that fires up 1 megawatt of a coal plant, then you just added 1 megawatt or a certain footprint of carbon. And so it's very location specific, and it's a point in time specific, which is super important because when you're analyzing carbon emissions, you really want to take those factors into account because averages don't really tell you a lot. We wanted to do this independent analysis because we feel it's important to validate what we've always believed about our facilities and that placing them with renewables, consuming wasted energy would create a completely new standard for data centers. And we're happy to announce that the independent analysis found that compared to our data center counterparts in ERCOT in the West territory of the state. Project Dorothy emits 18% less carbon annually. And it should be noted that ERCOT West is a renewable heavy region, and so it has a lower LME rate than ERCOT overall. And that's something we are consciously doing, putting our facilities where there's more green electrons and more importantly, more wasted green electrons. And by doing that and a number of other things that we do in our design, it allows us to be almost 20% less carbon intensive than other debated centers.

Got it. There was also a figure in the report about 40% cleaner. Can you explain the difference between those two?

Right. So the first number, 18% is the sort of headline of the announcement is the LME method. So that's the method I just talked about. It measures our footprint at that location. And again, the report found that we emit 18% less carbon than traditional around the clock data centers located in West Texas. So if they take a data center that might be running a Netflix stream or something like that or other types of high-performance applications. If we did the same thing in our facility or if that data center was doing AI, if we did the same thing, we would emit 20% less carbon, give or take. Now the 40% is related to a study that came out sometime last year that was highly publicized by New York Times. It was a New York Times study that was assisted by WattTime, which is a similar company to REsurety. We chose REsurety in our analysis because they've got a way more advanced data set, much higher granularity in terms of location. I think WattTime has probably 7 or 8 locations in terms of zones that you can be in, REsurety has thousands of them. And so they really have a more robust data set. What we did is we went back to that New York Times WattTime and just for everybody's benefit, what the WattTime study said was that, hey, you carbon data center -- you crypto data centers out there are actually not helping the environment. You actually forcing more carbon to be emitted. And here's a stack ranking of all of the different data center counterparts. What we did is we looked at that report and it found that a sort of an average megawatt in their sample set, induced emissions of call it, 4,188 marginal [ carbon ] tons per year. So take an average megawatt being produced by or is consumed by these crypto places based on their location and whether the power was there, call it, 4,000 CO2 tons per year. Now what we did is we took a similar methodology. We used the same methodology, but then we used our location in the analysis and reran the model. And what we found that due to Dorothy's location in that renewable rich part of Texas, and our ability, which is super important to use the curtailed wind energy induced an emissions level of 2,657 carbon tons per year. That's almost 40% less than our data center counterparts. And that's what that second number was meant to portray that depending on how you look at the analysis comparing to traditional data centers compared to crypto using the WattTime analysis, we're significantly less, and that is a -- in our view, a pivotal point for us because not only have we proven we can build these behind the meter models, develop partnerships with renewable energy power plant partners we can execute and build the facilities and get them monetizing that energy, but we're actually doing something better for the environment, which has broad implications, you asked me what's the -- what's the significance of this. One, it proves the Soluna business model works; two, it positions us well for AI, Graham. There's lots of talk about the growth of AI and what's going to be happening around that compute. And we believe that what we've done here with Dorothy on the crypto side is a great blueprint. It's a better mousetrap for the future of computing. And here's why. AI has 3 key challenges that no one is really talking about right now. We talk a lot about the carbon emissions, the fact that crypto was bad for the environment and so forth. But we don't really talk a lot about traditional data centers out there, how much energy is consumed, et cetera. And we're now talking a lot more about data centers and compute because of the rise of AI. AI has 3 key challenges. One is that it's really hungry. So if you look at the power density of AI compute compared to other high-performance compute that might be sitting inside of a traditional data center. It's energy demand is exploding. If you look at a typical rack, and I had to explain this to someone yesterday, if you walked into a data center and you look at a rack, you've got the raise floors and everything, that rack consumes somewhere on about 7 to 8 kilowatts for that rack. You've got a bunch of machines and CPUs running all sorts of software. If you were to put AI in that same space, it would demand 10x the amount of energy, somewhere around 50 watts to 60 watts of energy -- i meant to say kilowatts of -- 7 kilowatts to 8 kilowatts versus 50 kilowatts that's 10x and so if you look at that data center all of a sudden you need a lot more power and you need a lot more space to crunch in that much energy demand and anecdotally there is a view that AI will consume over the next decade more power than the entire global to data center space alone, there are some sort of projections that I've heard sort of reaching 20 gigawatts to 30 gigawatts. So if you're tracking with me so far, Graham, AI is hungry, right? The second, it's thirsty, actually, traditional data centers particularly use a lot of water and water consumption to drive the heat rejection and cooling management and these chips get really hot, very similar to crypto. Microsoft use an equivalent estimate of about 2.8 million glasses of water to train ChatGPT. And that's due to this concept where heat exits the building. It goes into a cooling tower, and they use evaporation to cool that. And so there's water being pulled out of the ground and evaporating, that's a high carbon footprint, right? That's a high issue from a scope 1 and 2 perspective. And then the last thing is AI is dirty. So it's -- these facilities tend to be based in population centers. Most hyperscalers are in Virginia, Arizona or these places where you've got big access to bandwidth closer to the use of the compute. But if you look at data centers on a global basis and the U.S. greenhouse gas emissions, if you will, 2% of the emissions are due to traditional data centers. On a global basis, power consumption and greenhouse emissions is a pretty significant as it relates to data centers and over half of that emission is coming from the U.S. alone because we have half of the data center footprint on a global basis. If you look at GPT, Gopher, Bloom, OPT, they are consuming somewhere around 900 tonnes of carbon emissions, these are big numbers. And so what we're doing here can be applied to addressing these issues by building a new type of data center that's based completely in a completely different location that could create a more sustainable, scalable AI platform. I know that's a long answer to your question, Graham, but we've been thinking about this a lot, and we're pretty excited about this.

No, that is great. And I mean it's -- I think very -- I've seen from the beginning, it's unfair that the New York Times painted everybody with the same brush because not all data centers are the same. And the locations are different, and it's kind of intellectually I'll say it. I think it's intellectually dishonest to try and not look at them differently because it is different if you're in a low -- high renewable place versus not. So I mean, it's great that you were able to do that and you report and actually separate the difference and show that. But on that issue about the power demand, I mean, that's a considerable issue that with this country and the world is going to have to face because the power grid adequacy of reserve margins are getting tight. We've seen what's happened over the last few years and the heat at the peaky levels, there's not enough power there. It's a problem that's going to have to be solved and nobody wants to build large baseload power plants.

That's right. Yes.

So the solution is going to have to be found around it. And that's where demand response is probably one of the biggest things that are out there, which ties into how you set up your data centers.

Exactly right. Most data centers -- traditional data centers are around the clock facilities. They are not resources for the grid. We're using our -- sort of better mousetrap, if you will, can achieve greater efficiency through then our ERCOT counterparts because of our proprietary Maestro OS software, the fact that we can utilize the advanced aerodynamics and thermal dynamics of our facilities to cool the facility without using water. The design enables the energy consumption by the facilities that almost entirely goes towards powering the computers and the operations. And more importantly, we can serve as a demand response solution for the grid. So imagine our entire facilities are full of AI compute, let's say, and so we're delivering massive compute capability to the marketplace as the demand rises our carbon footprint is significantly less than a traditional data center and our flexibility powered by our proprietary technology then allows us to deliver a resource to the grid that it can use in emergency situations where it needs more power, and it needs to be able to bring more of that green energy to the grid to power situations where as we've seen with some of these cold snaps or the heat the heat explosions in the summer. When you put all of that together, we're creating a new type of data center that enables greater renewable energy integration sustainable computing and a utility scale demand response solution, which helps shape the future of the grid and the future of computing.

Yes, because ultimately, when you need more power, you are, they put on more megawatts, which is going to be challenged or you do the megawatts, just take it away, which is demand response.

Exactly.